News for hackers. Stuff that matters.
Re:Will the Fear of Being Confused for AI Mean That We Will ... by ltbarcly3 on Sunday July 20, 2025 @06:20AM (#44623647)
AI writing will be impossible to distinguish from human writing in a year or two. The weird crappy way it writes now will be adversarially driven out of it in the next LLM releases, and probably gone in one or two more.

No more lists of 3 things, no more emdash, no more vacant live laugh love level vapid niceties.

Re:Will the Fear of Being Confused for AI Mean That We Will ... by janice1999 on Sunday July 20, 2025 @07:22AM (#44624098)
I don't think or at least hope so. The corporate desire to sanitise LLMs for "safety" (for PR/liability) and making it agreeable (for engagement) will hopefully leave AI with a recognisable non-human style.
Re:Will the Fear of Being Confused for AI Mean That We Will ... by bryanrasmussen on Sunday July 20, 2025 @07:15AM (#44624042)
>No more lists of 3 things, no more emdash, no more vacant live laugh love level vapid niceties.

So only humans will have these things then?

Re:Will the Fear of Being Confused for AI Mean That We Will ... by JKCalhoun on Sunday July 20, 2025 @08:09AM (#44624418)
Just realtors; realtors describing properties they are listing.
Re:Will the Fear of Being Confused for AI Mean That We Will ... by dukeyukey on Sunday July 20, 2025 @06:33AM (#44623736)
AIs will still have a "default" style that most people will leave because people are lazy, and that default style will be spottable.
Re:Will the Fear of Being Confused for AI Mean That We Will ... by russfink on Sunday July 20, 2025 @06:59AM (#44623916)
I wish I knew how to (ask an AI to) characterize this style. Is it always three examples? Nice formatting and bullet points? Evenly issued, lengthy and didactic explanations?
Re:Will the Fear of Being Confused for AI Mean That We Will ... by scarface_74 on Sunday July 20, 2025 @07:58AM (#44624333)
Following the guidance of another article that was just posted here, I’m going to just share a ChatGPT conversation…

https://chatgpt.com/share/687cd655-1cd8-8010-926d-645cefe928...

When I write, I now go out of my way not to use lists or em dashes. I learned before LLMs, that using lists in writing was lazy. Even when I do use an LLM for writing, I tell it to explain everything in paragraph form. I start off with my own outline and then “discuss” what I want to emphasize.

https://brevity.wordpress.com/2014/04/21/10-reasons/

The second thing about LLM generates text especially when doing technical writing, is that anytime it explains something, even with a lot of context of what I’m working on, it always adds lines about “benefits”.

In my use case something like “this provides a secure, adaptable environment…”. I have had to remove wording I had like that on my resume before LLMs and definitely don’t put it in my own writing now.

Ironically enough, out of the four examples ChatGPT generated about “what makes a good leader” in the link above, the example of “AI generated”, is the one I would lean toward in my own writing.

Also, putting AI assisted writing back through a new session with ChatGPT and asking it does it sound AI generated multiple times, and taking its suggestions will make it sound less like it’s AI generated.

When I do write something “thought leader”ish which unfortunately is part of my job now as a staff architect, I give a lot of real world examples that couldn’t easily be fabricated by AI and lean on those examples as I make my points.